This study aims to analyze viscoelastic properties of the wrist in patients with Parkinson's disease (PD) in comparison with the clinical score of severity. Forty-five patients with PD and 12 healthy volunteers participated in this study. Severity of rigidity at the wrist was rated by a neurologist just before the experiment. Wrist joint torque resistive to the imposed movement was measured. Three different models, (identical in structure, only different in the number of parameters for extension and flexion phases) were used in identification of viscoelastic properties: 1) one damping constant and one spring constant throughout all phases, 2) two damping constants for each phase and one spring constant throughout all phases, and 3) two damping constants and two spring constants for each phase. Normalized work and impulse suggested in the literature were also calculated. Spring constants of different models and phases showed comparable correlation with rigidity score ( r=0.68-0.73). In terms of the correlation of damping constant with clinical rigidity score, model 1 ( r = 0.90) was better than models 2 and 3 ( r=0.59 - 0.71). These results suggest that the clinical rigidity score is better represented by the mean viscosity during both flexion and extension. In models with two dampers (model 2 and 3), the damping constant was greater during extension than flexion in patients , in contrast that there was no phase difference in normal subjects. This suggests that in contrast with normal subjects, phase-dependent viscosity may be an inherent feature of PD. Although work and impulse were correlated with clinical rigidity score ( r = 0.11 - 0.84), they could not represent the phase-dependent rigidity inherent in PD. In conclusion, the viscosity of model 1 would be appropriate for quantification of clinical ratings of rigidity and that of model 2 for distinction of PD and also for investigation of phase-dependent characteristics in parkinsonian rigidity.